Appendix 1 - Identifying the Complex Aspects of a System.docx

Appendix 1 to BAA FA8750-21-S-7005

Identifying the Complex Aspects of a System

This document is organized into two major sections: properties of the system and actions on the system. Each of these sections includes subsections and characteristics that describe how the system incorporates specific aspects of complexity.

Each characteristic is examined from two perspectives: inward looking and outward looking. Inward looking indicators measure the extent to which the system incorporates that characteristic into itself, as a structural property or as a defensive feature. Outward looking indicators measure how much the system makes use of the given feature, either exploiting its presence in the adversary or as a means of attacking the adversary.

Framing a System in CAS Terms To understand the extent to which a given system is or is not complex, it is essential to frame it in appropriate terms. We have found that it is often extremely useful to think of the system as a network of some sort, potentially with many interacting parts and internal feedback loops. By definition, any complex system can be framed in this way. In contrast, a simple or even complicated system will correspond to a simpler network with only limited internal feedbacks. For example, a diagram of how a rifle works will be a set of linear steps, with few if any feedbacks or parallel processes. On the other hand, a diagram of how an air campaign is conducted can easily be drawn as a complex web of people and equipment that are continually adjusting and adapting to one another. The rifle is a simple system, the air campaign is a complex one.

The relevant network might be in many dimensions including social, geographical, temporal, functional, and so forth. As such, it is critical that the analyst think through the various ways that the system under analysis might be framed in these network terms.

With the system conceived in network terms, we divide its analysis into two parts:

1. To what extent is it organized as a complex adaptive system (CAS)?

2. To what extent does it make use of CAS principles?

In the sections that follow, we unpack CAS properties and actions by describing them abstractly, and pointing out how they might be relevant to Air Force concerns.

Properties of the System This first section of the rubric asks the analyst to think about the structure and properties of the network implied by the system under analysis. The characteristics in this section are chosen to provide a minimal sufficient description of the properties of CAS in general. Conceiving of the system in these terms is the first step in viewing it through the CAS lens. We have kept this conception extremely simple to allow for intuitive application.

Organization The first set of metrics look at how the system is organized. Here we are thinking of the structure of the network itself in its various dimensions.

Network To what extent does the system involve a network structure? Is it a simple network or a complicated one?

· Inward looking: How much is the system itself designed and constructed as a network?

· Outward looking: How much is the system designed to exploit or attack the adversary network structure.

Multi-Level To what extent does the system involve multi-level, nested network structures? These may include modular components that have complex internal connections and simpler connections to other modular components (like the internal organ structure of a living thing, a decentralized computing network, or an interacting set of specialized military units).

· Inward looking: To what extent are such structures part of the system itself?

· Outward looking: To what extent is the system designed to exploit or attack multi-level or nested networks structure in the adversary? Providing an ability to identify and disable critical components might be an example of this.

Adaptation The next set of metrics look at the ways in which the system makes use of adaptation. Adaptation is a second defining feature of CAS and this adaptation is often decentralized.

Adaptive To what extent is adaptation a part of the system? Is there reaction to the environment, or does the system always do the same thing?

· Inward looking: Does the system itself adjust and adapt to changes in its environment (including adversary actions)?

· Outward looking: Is the system designed to disrupt or otherwise exploit adversary adaptation?

Self-organized To what extent does any adaptation have a bottom-up character? Are changes determined centrally, or is change decentralized, with subcomponents reacting and adapting independently to make the overall system a better fit for its environment?

· Inward looking: To what extent does the system adapt in a decentralized way?

· Outward looking: To what extent does the system exploit or disrupt the adversary’s ability to adapt in a decentralized way?

Temporal CAS adaptations generally do not happen all at once but take time to propagate through the system. The system adapts to change over time learning from experience.

· Inward looking: Does the system learn over time? To what extent do changes take time to propagate through the system?

· Outward looking: Does the system exploit or disrupt the adversary’s ability to learn over time? For example, is the system designed to learn more quickly than the adversary system?

Emergent Properties CAS tend to have properties at higher scales that cannot be predicted or understood in terms of the properties of the constituent components.

Phase transitions Does the system involve phase transitions? These would include abrupt changes in behavior that are not centrally determined, and potentially decisions that are taken as a result of information coming from dispersed parts of the system.

· Inward looking: Does the system change its behavior abruptly in response to more continuous changes in its environment?

· Outward looking: Does the system anticipate, preempt, or exploit abrupt changes in adversary behavior?

Domains of Operation Domains of operation do not emerge during the operation of a particular system but represent categories of equipment and responsibilities that emerge from operations over time. Land, sea, and air are the most basic of these, but at the same level we also have categories like space, cyber, information, etc. At lower levels other domains might be identified with the Air Force core missions of air and space superiority; intelligence, surveillance, and reconnaissance (ISR); rapid global mobility; global strike; and command and control (C2).

· Inward looking: Does the system operate across more than one domain of operation?

· Outward looking: Does the system operate on or disrupt connections between more than one adversary domain?

Other Emergence is a notoriously hard category to define objectively and to apply consistently. However, emergent phenomena are often critical to understanding the workings of a complex adaptive system. Pending further development, this encourages the analyst to think about other aspects of emergence that might be relevant to the system and assess them.

· Inward looking: Are there other examples of emergence that can be identified as important to the operation of the system?

· Outward looking: Is the system designed to interact with, attack, or disrupt other emergent features of the adversary system?

CAS Inspired Actions This second section asks the analyst to think about the actions that the system takes or defends against. It is intended to focus on the extent that the system is leveraging CAS principles in the actions it is designed to carry out. It is important to note that there are many ways for a system to be successful – this document is designed to highlight the extent to which CAS principles are central to the design of the system, not to evaluate its overall worth.

Exploiting Bounded Rationality In most complex adaptive systems, the component parts of the system (whether they be people or pieces of equipment) have neither perfect information about their adversary and environment, nor do they have perfect reasoning ability to process the information that they do have given the constraints of time, memory, etc.

Create Imperfect Information To what extent does the system operate to gain advantage from incomplete information?

· Inward looking: Is the system designed to protect blue information? Is the denial of information to the adversary a major design goal? (stealth, encryption, covert activity, etc.)

· Outward looking: Is the system designed to gather or obtain information? Is discovering information about the adversary a major design goal?

Insert False Information To what extent does the system deal with false information?

· Inward looking: Is the system designed to reject false information provided by the adversary?

· Outward looking: Is the system designed to provide false information to the adversary?

Instill Deep Uncertainty Deep uncertainty in complex systems means that the actor not only does not know important things but does not even know what needs to be known. The actor does not know the rules of the system and is not able to accurately anticipate the outcome of their own actions. Examples might include strategic shifts, randomized actions, or other measures designed to create an uncertain decision-making environment.

· Inward looking: Is the system specifically designed to maintain clarity and reduce the probability of deep uncertainty?

· Outward looking: Is the system specifically designed to keep the adversary uncertain about the workings, capabilities, and intentions of blue?

Span Boundaries Conceiving of both blue and red as networks suggests various methods that blue might use to gain advantage. The nested structure of these networks means that not all nodes have equal value and invites analysis of which interventions might be most effective in preserving blue capabilities and disrupting red capabilities. Disruption of some nodes may have impact or influence across different parts of the system. Some actions may have impacts on nodes that are not closely connected. This section explores ways that the system makes use of these properties of network structure.

Exploit leverage points Does the system take advantage of nodes that play particularly critical roles in system operation?

· Inward looking: Does the system include measures to minimize the criticality of any one of its nodes? Does it take particular measures to identify and make its critical nodes particularly robust?

· Outward looking: Is the system designed to identify and disrupt nodes in the adversary system that are particularly important to its operation? Is it designed to operate on points where a relatively small intervention can have a disproportionately large effect?

Exploit diffuse decisionmaking Many CAS involve decentralized decision making. This can provide robustness in the face of losses and disruption in communications but can also lead to inefficiency if these decisions cannot be well coordinated.

· Inward looking: Does the system incorporate diffused or decentralized decision making as part of its design?

· Outward looking: Is this system designed to take advantage of diffuse adversary decision making, for example, by disrupting adversary ability to coordinate decisions?

Maximize interdependencies The parts of a CAS are interdependent, and these interdependencies operate across a range of scales, distances, and levels of directness. These interdependent structures allow for specialization and flexibility; however, they can also produce weaknesses when a response depends on a chain of actions that is not sufficiently robust or resilient.

· Inward looking: Does the system have interdependent components? If so, are these interdependent parts designed to work together in a robust, resilient way?

· Outward looking: Does the system require the adversary to respond with multiple, interdependent systems?

Adapt Adaptation is an essential feature of complex adaptive systems. This section examines how the system incorporates adaptation in the face of changing circumstances.

Gather feedback In order to adapt to circumstances, the system needs to gather feedback on what is happening and what is resulting from its current actions.

· Inward looking: To what extent does the system monitor its own functioning?

· Outward looking: To what extent does the system gather information about the adversary to evaluate the impact of its previous interactions?

Make use of feedback Having gathered feedback, the system must make changes and adjustments in order to put that feedback into action in adapted behavior.

· Inward looking: How much is the system able to use information about its internal functioning to make adjustments and adapt itself to changing circumstances?

· Outward looking: How much is the system able to change its behavior in response to changes in adversary behavior or other external circumstances?

Leveraging Dynamic Non-linearities A side effect of the complex network structure, decentralized control, feedback, and adaptation of CAS is that they tend to contain areas of dynamic non-linearity – parts of the behavior space where small changes in input can produce large (or even discontinuous) changes in output.

Overload In many cases the performance of a system falls off quite abruptly as it reaches some sort of saturation point. An aircraft controller might be able to handle a large workload with few errors, but there comes a point where the controller is overloaded and the rate of errors increases dramatically.

· Inward looking: To what extent is the system robust to overload?

· Outward looking: To what extent is the system designed to create and take advantage of overload in the adversary?

Overwhelm Overwhelming takes place when one party perceives its disadvantage to its adversary to be so large that resistance is futile, pursuing instead strategies like retreat, surrender, or simply absorbing the blow.

· Inward looking: Is the system robust to being suddenly overwhelmed? Does it behave as well as it can be expected to do when it is outmatched?

· Outward looking: Is the system designed to present the adversary with overwhelming force (which need not be kinetic)? Does the system seek to make use of the concept of overwhelming the adversary?